joe@GENETICS.WASHINGTON.EDU (Joe Felsenstein) (12/04/90)
Sarah Barron asked if PHYLIP has been ported to Cray Y-MPs. It has, by Ron DeBry (debry@ds1.scri.fsu.edu) and by Shan Duncan (duncan@csd4.csd.uwm.edu) (the latter was actually an X-MP). Both succeeded in getting it to work without difficulty. But both reported sufficiently slow running times that it was obviously not making use of vector-processing very well. The problem is that some of the interior loops of the program use scalar quantities to store intermediate results. If these scalars were made into vectors, there would, I suspect, be a big speedup. Also, in some programs there are IF statements inside the innermost loops that can have results that come out differently for different values of the index variable. That would seem to be something that would not vectorize well but might respond better to true parallel processing. Programs that should do particularly well if the scalar temporary storage variables are made vectors are DNAML, DNAMLK, and RESTML. There are pointers in the main documentation file under Cray, and I will try to be responsive to inquiries. ----- Joe Felsenstein, Dept. of Genetics, Univ. of Washington, Seattle, WA 98195 Internet/ARPANet: joe@genetics.washington.edu (IP No. 128.208.128.1) Bitnet/EARN: felsenst@uwalocke UUCP: ... uw-beaver!evolution.genetics!joe